handle case of duplicate passages

spanish_passage_retrieval.py

@@ -106,12 +106,16 @@ class PRES(datasets.GeneratorBasedBuilder):
                     qrels_s2s[qid] = []
                     qrels_s2p[qid] = []
 
+            known_passage_ids = set()
+
             for annotated_topic in rel_passages['topics']:
                 topic = annotated_topic['number']
                 for annotation in annotated_topic['annotations']:
                     passage_id = f'doc_{annotation["docNo"]}_{annotation["start"]}_{annotation["end"]}'
                     doc_id = f'doc_{annotation["docNo"]}'
-                    corpus_sentences.append({'_id': passage_id, 'text': annotation['text']})
+                    if passage_id not in known_passage_ids:
+                        corpus_sentences.append({'_id': passage_id, 'text': annotation['text']})
+                        known_passage_ids.add(passage_id)
                     for qid in topic_to_queries[topic]:
                         qrels_s2s[qid].append(passage_id)
                         qrels_s2p[qid].append(doc_id)